The field of clinical diagnosis and analysis is rapidly advancing with the integration of multimodal large language models (MLLMs) and traditional deep learning techniques. Recent developments have focused on improving the accuracy and efficiency of disease diagnosis, clinical risk prediction, and medical image analysis. Notably, the use of MLLMs has shown promise in handling complex clinical data, such as unstructured text and images, and in generating informative demonstrations for in-context learning. Furthermore, the application of explainable AI techniques has enhanced the transparency and trustworthiness of diagnostic models, allowing for better understanding of diagnostic uncertainties and improved clinical decision-making. The combination of these advances has the potential to revolutionize clinical practice, enabling more accurate and efficient diagnosis, and ultimately improving patient outcomes. Noteworthy papers in this area include the proposal of Retrieval-Augmented In-Context Learning (RAICL) framework, which dynamically retrieves informative demonstrations to enhance in-context learning in MLLMs, and the development of ConfiDx, an uncertainty-aware large language model that excels in identifying diagnostic uncertainties and generating trustworthy explanations for diagnoses.